Verbal particles in Swedish such as `upp' (up) and `bort' (away) are defined by the following criteria:

     1) a particle is stressed
     2) a particle immediately follows the verbal position within the VP
     3) a particle cannot have a modifier or a complement

Although it is easy to tell particles apart from non-particles, it is not immediately obvious how they are formally different from other linguistic entities. I will show that they cannot be analyzed as a distinct syntactic category, on a par with nouns and verbs, for example. I will also show that they do not correspond to a unique grammatical function, which can be contrasted with subject and object. Instead, I propose that the unique characteristics of particles are due to the fact that they are phrase-structurally different from other words and phrases: particles are non-projecting words in the syntax.

The analysis will be cast in Lexical-Functional Grammar (LFG), which separates the phrase structural representation (c-structure) from other syntactic levels of representation (functional structure and argument structure). In order to accomodate for (as well as constrain) the presence of non-projecting words cross-linguistically, I will modify the LFG version of X'-theory.

My account of particles is very similar to many previous syntactic analyses of clitics. I will therefore revisit the notion of `clitichood', and show that the present view of particles leads to a new typology of clitics. I argue that words are classified along two dimensions: phonological dependence and syntactic projectivity. Prototypical clitics are non-projecting and phonologically dependent, whereas Swedish particles are non-projecting and phonologically independent.